Neutrophil Gelatinase–associated Lipocalin Predicts Short-term Outcomes in Decompensated Cirrhosis With Acute Kidney Injury

Abstract

Background and aims

Acute kidney injury (AKI) increases mortality in cirrhosis. Early identification of the cause of AKI helps in planning appropriate management. We aimed to find whether neutrophil gelatinase–associated lipocalin (NGAL) can be used to differentiate between different types of AKI in cirrhosis and predict short-term outcomes in patients with decompensated cirrhosis and AKI.

Method

This was a time-bound study in which consecutive hospitalized patients with cirrhosis and AKI were prospectively recruited and managed as per standard care. Acute on chronic liver failure (ACLF) was diagnosed as per the EASL-CLIF Acute-on-Chronic Liver Failure in Cirrhosis (CANONIC) criteria. Urine NGAL was measured by enzyme-linked immunosorbent assay (ELISA) by Epitope Diagnostics Inc. kit (San Diego, USA.) in all patients on admission, and patients were followed up until hospital discharge or death.

Results

A total of 110 consecutive patients (median [range] age: 44 [28–81] years;87.3%were male; ACLF: 71.8%; acute decompensation 28.2%; Model for end-stage liver disease (MELD) 27 [13–46]; Child-Turcotte-Pugh (CTP) 11 [7–15]) with cirrhosis and AKI were recruited. Alcohol was the most common etiology of cirrhosis(64.5%)). Pre-renal azotemia (PRA) was the most common cause of AKI (n = 56). Urine NGAL was significantly elevated in acute tubular necrosis (ATN) (1747 [6–6141] ng/ml than in hepatorenal syndrome (HRS) (379 [33.5–2320] ng/ml; P < 0.0001) and PRA (167 ng/ml [3.34–660]; P < 0.0001). Sixty-four percent patients with ATN, 27.6% patients with HRS, and none with PRA required dialysis. A total of 79.31% patients with HRS and 76% with ATN died. Urine NGAL was significantly higher in patients who required hemodialysis than in those who did not (1733 [243–6141] ng/ml vs 235 [3.34–2320] ng/ml; P < 0.0001). Both urine NGAL (n = 110) and plasma NGAL (n = 90) were significantly higher in patients who died (urine NGAL: -475 [6–6141] ng/ml vs 247 [3.34–2320] ng/ml; P = 0.002;plasma NGAL-950 [94–4859] ng/ml vs 608 [18–3300)]g/ml; P < 0.001). On multivariate analysis, urine NGAL and INR could predict mortality.

Conclusion

NGAL can differentiate between different types of AKI in cirrhosis and predict the need for hemodialysis and mortality in decompensated cirrhosis with AKI.

Graphical abstract

One hundred-ten patients (median age: 44 years) with cirrhosis and acute kidney injury (AKI) were prospectively recruited. Alcohol was the most common cause of cirrhosis and acute decompensation in acute on chronic liver failure (n = 79, defined by CANONIC criteria). Urine and blood neutrophil gelatinase–associated lipocalin (NGAL) was measured by ELISA by Epitope Diagnostics Inc. kit (San Diego, USA). Urine NGAL was significantly elevated in acute tubular necrosis (ATN) (1747 [6–6141] ng/ml than in hepatorenal syndrome (HRS) (379 [33.5–2320] ng/ml; P < 0.0001) and pre-renal azotemia (167 ng/ml [3.34–660]); P < 0.0001). Urine NGAL was significantly higher in patients who required hemodialysis than those who did not [(1733 [243–6141] ng/ml vs 235 [3.34–2320] ng/ml; P < 0.0001). A total of 52 patients died (HRS: 23, ATN: 19 and Prerenal azotemia: 10). Both urine NGAL (n = 110) and plasma NGAL (n = 90) were also significantly higher in patients who died. Urine NGAL may be incorporated into routine patient care to predict short-term outcome in patients with cirrhosis and AKI.

Acute kidney injury (AKI) occurs in 15–25% of hospitalized patients with cirrhosis and is associated with high mortality.^{1, 2, 3} The causes of AKI in cirrhosis include pre-renal azotemia (PRA), hepatorenal syndrome (HRS) and acute tubular necrosis (ATN).⁴ Of these, the occurrence of HRS is associated with a 1-year mortality of up to 60%.⁵ The diagnosis of HRS requires the exclusion of PRA by plasma volume expansion for 48 h, thereby delaying the initiation of vasoconstrictors. The differentiation of HRS and ATN is equally important as volume expansion and vasoconstrictors, which are the mainstay of treatment in HRS, may be futile or deleterious in ATN.⁶

Serum creatinine, which is most commonly used to assess renal function, is less reliable in cirrhosis detection as its level is reduced due to its decreased hepatic formation, decreased secretion due to spironolactone, low muscle mass in cirrhosis, and high bilirubin levels, which impairs its accurate measurement by spectrophotometry.⁷ It also has no role in differentiating between the various causes of AKI. Various serum and urine biomarkers have been studied for their ability to differentiate between functional and structural forms of AKI. Of these, while Kidney injury molecule-1 (KIM-1) and NGAL are renal tubular proteins upregulated in tubular injury, Cystatin C is produced by all nucleated cells and is filtered by the glomerulus. In renal injury, Cystatin C levels rise due to decreased reabsorption and catabolism by proximal tubule.⁸ NGAL is the most widely studied biomarker in AKI, and although considerable literature exists to support its utility, NGAL estimation is not yet routinely recommended by guidelines.⁹ Its role as a prognostic marker in cirrhosis and the need for hemodialysis as well as in ACLF is less well studied.

In this study, we assessed the ability of NGAL to differentiate between structural and functional renal injury as well as its prognostic significance in patients with decompensated cirrhosis and AKI.

Methods

Population and Setting

This was a prospective time-bound study, where consecutive patients hospitalized in our institute with cirrhosis and AKI between January 2020 and November 2021 were recruited after obtaining informed consent. Patients with human immunodeficiency virus infection, hepatocellular carcinoma, pregnancy, lactation, obstructive uropathy, and chronic kidney disease were excluded. All study patients were evaluated and treated as per the standard of care. Based on their evaluation, they were further classified into PRA, HRS, and ATN.

NGAL Estimation

NGAL levels were estimated in urine in all patients (uNGAL) and also in plasma in a subset of patients (pNGAL) at admission. Five to 10 ml of urine was collected for measurement of uNGAL, and 3 ml of blood was collected in an Ethylenediaminetetraacetic acid (EDTA) vial for measurement of pNGAL. The samples were centrifuged at 4 °C at 1500 g for 15 min. Plasma was separated in cases where blood was collected. Plasma and urine NGAL were measured in supernatant by ELISA using Epitope Diagnostic Inc. kits (San Diego, USA).

Clinical and biochemical parameters, disease severity scores, uNGAL levels, pNGAL levels, and clinical outcomes were compared between the above three subsets of AKI patients.

Definitions

ACLF was defined according to the European Association for the Study of the Liver (EASL) criteria.¹⁰ AKI was defined by an increase in serum creatinine of 0.3 mg/dL within 48 h or to 1.5-times that of the baseline within seven days.¹¹ Pre-renal azotemia was defined as AKI (a 50% increase in the serum creatinine level from the baseline that is known or presumed to have occurred within the past 7 days or a rise of 0.3 mg/dl in serum creatinine level in <48 h or decrease in urine output <0.5 ml/kg/hour for 6 h) that responded to 48 h of volume expansion with albumin at a dose of 1 gm/kg (maximum up to 100 gm) if creatinine was more than 1.5 mg/dl or by 0.9% normal saline if creatinine was less than 1.5 mg/dl. HRS was defined as patients with AKI who fulfill the following criteria: 1) cirrhosis with ascites, 2) sCr >1.5 mg/dL, 3) no improvement of serum creatinine (decrease to a level of <1.5 mg/dL) after at least 2 days with diuretic withdrawal and volume expansion with albumin, 4) absence of shock, 5) no current or recent treatment with nephrotoxic drugs, and 6) absence of parenchymal kidney disease as indicated by proteinuria >500 mg/day, hematuria (>50 red blood cells per high power field), and/or abnormal renal ultrasonography.¹² ATN is a clinicopathological syndrome characterized by intrinsic renal injury secondary to ischemia, sepsis, or toxin. It is confirmed by renal biopsy.^13,14 However, our patients carried a high risk of complications if subjected to renal biopsy. So we used a composite of clinical and laboratory criteria. ATN was defined by the fulfillment of at least three out of four of the following criteria: FeNa > 2%, urinary osmolality < 400 mOsm/L, urinary sodium level > 40 mEq/L, and presence of shock or use of nephrotoxic drugs.¹⁵ AKI was further staged as stage 1—a rise in the serum creatinine level of 0.3 mg/dl or an increase in serum creatinine ≥1.5-fold to 2-fold above the baseline; stage 2—an increase in serum creatinine level of >2-fold to 3-fold above the baseline; stage 3— an increase in serum creatinine level of >3-fold above the baseline or a serum creatinine level of 4 mg/dl with an acute increase of 0.3 mg/dl or initiation of renal replacement therapy. Progression of AKI was defined as progression of AKI to a higher stage and/or need of renal replacement therapy, and regression of AKI was defined as regression of AKI to a lower stage. MELD (Model for end-stage liver disease) score was calculated in all patients and chronic liver failure–sequential organ failure assessment (CLIF-SOFA) score in the subset, with ACLF as described in the respective studies.^16,17

Outcomes

The primary outcome of the study was the difference in uNGAL levels between patients with PRA, HRS, and ATN. Secondary outcomes were difference in pNGAL levels between the aforementioned three groups of AKI, diagnostic accuracy of NGAL in predicting mortality, and the need for hemodialysis.

Patients were followed up until discharge from the hospital. For the purpose of analysis, patients who were discharged in a terminal state were considered as dead.

Statistical Analysis

Normality of data were assessed by Shapiro–Wilk test. Continuous variables were expressed as median and range, whereas categorical variables were expressed as numbers and percentages. The groups were compared using Mann–Whitney test for nonparametric continuous variables and t-test for parametric continuous variables. The categorical variables were compared using the Chi square test. Correlation between variables was done by Spearman's coefficient. Diagnostic accuracy of NGAL to predict ATN, need for renal replacement therapy, and in-hospital mortality were studied using receiver operator characteristic (ROC) curve. A P-value of <0.05 was considered significant.

The study was approved by the institutional ethics committee (Letter number-AIIMS/IEC/20/595).

Results

A total of 110 patients with cirrhosis and AKI who fulfilled the aforementioned inclusion criteria were recruited to the study as shown in Figure 1. Of these, 56 (50.9%) patients had PRA, 29 (26.4%) patients had HRS, and 25 (22.7%) patients had ATN. Most of the patients with PRA had stage 1 AKI, whereas in both HRS and ATN, the AKI was of higher stage. The details of the AKI stage are given in supplementary figure S2.

Figure 1.

Flowchart showing the number of patients included and excluded.

The baseline characteristics of the study population are shown in Table 1. Of the 110 patients, 79 (71.8%) patients had ACLF, whereas the rest were admitted with acute decompensation (AD). Alcohol was the most common etiology of cirrhosis in the entire study population as well as that of acute hepatic insult in patients with ACLF. Twelve patients (10.9%) had diabetes mellitus, and 7 (6.4%) patients were hypertensive. While most of the parameters were comparable, blood urea, serum creatinine, serum bilirubin, and encephalopathy were significantly higher in the ATN group.

Table 1.

Baseline Characteristics of 110 Patients With Cirrhosis and Acute Kidney Injury.

Parametersa	Total (n = 110)	PRA (N = 56)	HRS (n = 29)	ATN (n = 25)
Age (years)	44 (24–81)	43.5 (26–81)	48 (24–70)	42 (28–74)
Sex (male)	96 (87.3%)	50 (89.3%)	24 (82.8%)	22 (88.0%)
Comorbidities
Diabetes mellitus	12 (10.9%)	6 (10.7%)	4 (13.8%)	2 (8%)
Hypertension	7 (6.4%)	3 (5.36%)	1 (3.4%)	3 (12%)
Etiology of cirrhosis
Alcohol	71 (64.5%)	32 (57.1%)	22 (75.9%)	17 (68%)
NASH	6 (5.5%)	2 (3.6%)	2 (6.9%)	2 (8%)
Chronic viral hepatitis	18 (16.4%)	14 (25%)	3 (10.3%)	1 (4%)
Others	15 (13.63%)	8 (14.28%)	2 (6.9%)	5 (20%)
Number of patients with ACLF	79 (71.8%)	29 (51.8%)	27 (93.1%)	23 (92%)
Etiology of acute insult in patients with ACLF
Alcohol	30 (37.9%)	7 (24.13%)	12 (44.44%)	11 (47.8%)
Infection	15 (18.9%)	8 (27.58%)	4 (14.81%)	3 (13.04%)
Viral	9 (11.39%)	5 (17.24%)	3 (11.1%)	1 (4.3%)
Others	25 (31.64%)	9 (31.03%)	8 (29.62%)	8 (34.78%)

ACLF—acute on chronic liver failure, AIH—autoimmune hepatitis, ATN—acute tubular necrosis, AVH—acute viral hepatitis, NASH—nonalcoholic steatohepatitis, PRA—pre-renal azotemia, HRS—hepatorenal syndrome.

^aContinuous variables expressed as median (range) and categorical variables as number (percentages)

NGAL Levels Were Higher in Patients With Severe Liver Disease and Poor Outcome

The details regarding disease severity are shown in Table 2. MELD, CLIF-SOFA score, and Chronic liver failure-Acute on chronic liver failure (CLIF C ACLF) scores were more in the ATN group than in HRS and PRA groups. Hemodialysis was required in 21.8% of the patients and was significantly more in the subset with ATN (64%) than in the other etiologies of AKI. However, mortality was similar in the HRS group and the ATN group (79.31% vs 76%; P = 0.51).

Table 2.

Biochemical Parameters and Disease Severity on Admission in 110 Patients With Cirrhosis and Acute Kidney Injury.

Parameters	Total (n = 110)	PRA (n = 56)	HRS (n = 29)	ATN (n = 25)	∗P value	∗∗P value	∗∗∗P value
Ascites	108 (98.18%)	54 (96.42%)	29 (100%)	25 (100%)	0.306	1	0.342
SBP	35 (31.8%)	18 (32.1%)	10 (34.5%)	7 (28%)	0.507	0.415	0.460
HE	75 (68.2%)	29 (51.8%)	21 (72.4%)	25 (100%)	0.054	0.004	0.000
UGI bleed	19 (17.3%)	11 (19.6%)	3 (10.3%)	5 (20%)	0.219	0.270	0.595
Hydrothorax	4 (3.6%)	4 (7.1%)	0	0	0.181	NA	0.221
STI	3 (2.7%)	2 (3.6%)	1 (3.4%)	0	0.733	0.537	0.475
HPS	1 (0.9%)	0	1 (3.4%)	0	0.341	0.537	NA
Hb (g/dl)	9 (3–14)	8.5 (3–12)	9 (3–12)	9 (6.2–14)	0.387	0.807	0.199
TLC (∗103/mm3)	9 (2.26–30)	6.8 (2.26–23)	11 (2.5–26)	15 (3.2–30)	0.001	0.118	0.000
Platelet (∗109/L)	102 (16–433)	97.50 (20–236)	114 (30–433)	90 (16–239)	0.559	0.487	0.935
Urea (mg/dl)	92.5 (6–332)	76 (6–225)	110 (30–255)	118 (25–332)	0.153	0.165	0.001
Creatinine (mg/dl)	2.1 (0.44–8.76)	1.68 (0.44–4.60)	2.4 (1.7–5.77)	3.57 (0.78–8.76)	0.00	0.034	0.000
Final Creatinine (mg/dl)	1.18 (0.33–5.57)	0.975 (0.33–1.53)	2.01 (0.79–4.94)	2.4 (0.63–5.57)	0.00	0.327	0.000
Bilirubin (mg/dl)	7.05 (0.53–44)	3.88 (0.53–42)	18 (0.7–44)	15 (1.24–36)	0.005	0.567	0.008
Albumin (g/dl)	2.5 (1.4–3.90)	2.65 (1.56–3.2)	2.5 (1.49–3.44)	2.2 (1.4–3.9)	0.461	0.285	0.09
INR	1.86 (1.06–4.5)	1.74 (1.06–4.5)	2.17 (1.2–4.03)	2.24 (1.06–3.17)	0.069	0.965	0.051
u NGAL (ng/ml)	372 (3.34–6141)	167 (3.34–660)	379 (33.5–2320)	1747 (6–6141)	0.0001	0.0001	0.0001
p NGAL (ng/ml) (n = 90)	703.5 (18–4859) (n = 90)	439 (18–1645) (n = 45)	780 (113.2–4859) (n = 22)	1989 (604–4057) (n = 23)	0.006	0.001	0.0001
FeNa (%)	0.6 (0.1–6.3)	0.5 (0.1–1.5)	0.4 (0.1–1.4)	2.2 (0.9–6.3)	0.161	<0.0001	<0.0001
Urinary osmolality (mOsm/l)	454.5 (277–864)	470 (298–813)	576 (389–864)	335 (277–497)	0.007	<0.0001	<0.0001
Urinary sodium (meq/l)	27 (7–132)	22 (9–47)	20 (7–42)	52 (33–132)	0.109	<0.0001	<0.0001
Proteinuria (mg/d)	147.5 (11–798)	125.5 (12–560)	128 (11–406)	297 (16–798)	0.767	<0.017	0.021
MELD	27 (13–46)	23 (13–46)	33 (17–44)	34 (14–45)	<0.001	0.481	<0.001
CTP score	11 (7–15)	10.5 (8–15)	12 (7–15)	13 (8–15)	0.006	0.153	<0.001
CLIF C ACLFa	47 (27–67) (n = 79)	42 (27–52) (n = 29)	49 (30–65) (n = 27)	55 (28–67) (n = 23)	0.000	0.014	0.000
CLIF SOFAa	10 (7–16) (n = 79)	9 (7–12) (n = 29)	10 (7–14) (n = 27)	12 (8–16) (n = 23)	0.028	0.015	0.000

∗-P-value for PRA vs HRS.

∗∗-P-value for HRS vs ATN.

∗∗∗-P-value for PRA vs ATN.

PRA—pre-renal azotemia, HRS—hepatorenal syndrome, ATN—acute tubular necrosis, CLIF SOF—Achronic liver failure–sequential organ failure assessment, SBP—Spontaneous bacterial peritonitis, HE—Hepatic encephlopathy, UGI—Upper gastrointestinal, STI—Soft tissue infection, HPS—Hepatopulmonary syndrome, TLC—Total leucocyte count, INR—International normalized ratio, MELD—Model for end-stage liver disease, CTP—Child-Turcotte-Pugh, uNGAL—urine neutrophil gelatinase–associated lipocalin, pNGAL—plasma neutrophil gelatinase–associated lipocalin, CLIF C ACLF—Chronic liver failure-Acute on chronic liver failure

^aCalculated only in patients with ACLF.

uNGAL showed moderate correlation with MELD score (r_s = 0.510; P < 0.001), Child-Turcotte-Pugh (CTP) score (r_s = 0.467; P < 0.001), and serum creatinine (r_s = 0.451; P < 0.001) at admission but weak correlation with CLIF-SOFA scores (n = 79; r_s = 0.362; P = 0.001).

pNGAL showed moderate correlation with CTP score (r_s = 0.407; P < 0.001) and weak correlation with serum creatinine (r_s = 0.395; P < 0.001), MELD score (r_s = 0.341; P = 0.001), and CLIF-SOFA score at admission (r_s = 0.237; P = 0.055).

NGAL Levels Had Higher Diagnostic Accuracy Than Creatinine in Differentiating ATN From Non-ATN Causes of AKI

Median uNGAL levels were the lowest in patients with PRA and highest in those with ATN (PRA vs HRS-167 [3.34–660] ng/ml vs 379 [33.5–2320] ng/ml, P < 0.001;HRS vs ATN- 379 [33.5–2320] ng/ml vs 1747 [6–6141] ng/ml, P < 0.001;PRA vs ATN-167 [3.34–660] ng/ml vs 1747 [6–6141] ng/ml,P < 0.001). Figure 2 shows that uNGAL could differentiate ATN from non-ATN type of AKI with a higher area under curve (AUC) than creatinine (0.879 vs 0.763).

Figure 2.

Receiver operating curve of neutrophil gelatinase–associated lipocalin and Creatinine for differentiation acute tubular necrosis and non–acute tubular necrosis type of acute kidney injury.

Predictive Value of uNGAL and pNGAL in Predicting Need For Hemodialysis and Mortality

uNGAL levels were significantly more in patients who required hemodialysis than in those who did not (1733 [243–6141] ng/ml vs 235 [3.34–2320] ng/ml; P < 0.001) and in the patients who died during admission than in the survivors (475 [6–6141] ng/ml vs 247 [3.34–2320] ng/ml; P = 0.002) as shown in Table 3. A total of 24 patients (HRS: 8 and ATN: 16) underwent hemodialysis for the indications of refractory hyperkalemia (n = 8), metabolic acidosis (n = 7), volume overload (n = 6), and uremic encephalopathy (n = 3). In majority of patients with PRA, the AKI regressed to a lower stage whereas in HRS and ATN majority of the patients progressed to a higher stage or remained in the same stage. The details are shown in supplementary figure S2.

Table 3.

Median Urine and Plasma Neutrophil Gelatinase–associated Lipocalin Value in Different Subgroups of Patients.

	uNGAL (n = 110)			pNGAL (n = 90)
Parameter	Yes	No	P-value	Yes	No	P-value
Hemodialysis (n = 24)	1733 ng/ml (243–6141) (n = 24)	235 ng/ml (3.34–2320) (n = 86)	<0.001	1933 ng/ml (113.2–4057) (n = 21)	611 ng/ml (18–4859) (n = 69)	<0.001
Death (n = 52)	475 ng/ml (6–6141) (n = 52)	247 ng/ml (3.34–2320) (n = 58)	0.002	950 ng/ml (94–4859) (n = 41)	608 ng/ml (18–3300) (n = 49)	<0.001

pNGAL—plasma neutrophil gelatinase–associated lipocalin, uNGAL—urine neutrophil gelatinase–associated lipocalin.

ROC curve for uNGAL, pNGAL, and serum creatinine for prediction of need for Hemodialysis (HD) is shown in Figure 3. AUC of uNGAL for prediction of need for HD was 0.885 (95% CI 0.807–0.963) and was higher than that for pNGAL (AUC: 0.821 [95% confifence interval {CI }: 0.714–0.928]) and serum creatinine (AUC: 0.806 [95% CI: 0.675–0.938]).

Figure 3.

Receiver operating curve of neutrophil gelatinase–associated lipocalin and admission creatinine for need for hemodialysis.

uNGAL at 503 ng/ml predicted the need for HD with a sensitivity of 81% and a specificity of 76.8%. pNGAL at 918.5 ng/ml predicted need for HD with a sensitivity of 76.2% and a specificity of 75.4%.

A total of 52 patients (47.3%) patients died in our study. Out of these, 23 patients had HRS, and 19 had ATN, but this difference was not statistically significant. Lowest mortality was seen in PRA (Table 4).

Table 4.

Clinical Outcomes in the Subgroups of Patients.

Parameters	Total (n = 110)	PRA (n = 56)	HRS (n = 29)	ATN (n = 25)	∗P value	∗∗P value	∗∗∗P value
Need for hemodialysis	24 (21.8%)	0	8 (27.6%)	16 (64%)	0.000	0.013	0.000
Duration of hospital stay (days)	8 (3–35)	9.5 (3–35)	6 (3–20)	9 (3–34)	0.005	0.072	0.794
Number of deaths	52 (47.3%)	10 (17.8%)	23 (79.31%)	19 (76%)	<0.001	1	<0.001

ATN—acute tubular necrosis, PRA—pre-renal azotemia, HRS—hepatorenal syndrome.

Both pNGAL and uNGAL could predict mortality in cirrhotic patients. The area under ROC for prediction of mortality was higher for pNGAL than for MELD score (0.704 vs 0.698) as shown in Figure 4.

Figure 4.

Receiver operating curve of neutrophil gelatinase–associated lipocalin, MELD, and CLIF C ACLF for mortality. CLIF C ACLF: Chronic liver failure-Acute on chronic liver failure.

On univariate analysis of uNGAL, presence of ACLF, serum bilirubin, and INR predicted mortality. However, on multivariate analysis, only urine NGAL and INR could predict mortality. (Table S2).

NGAL Levels in ACLF

Both uNGAL and pNGAL levels were higher in patients with ACLF than in patients without (Table S3, S4). Also, the trend of uNGAL levels being the highest in patients with ATN followed by HRS and lowest in patients with PRA was maintained even when the patients were stratified as those with and without ACLF. uNGAL predicted need for HD in ACLF with AUC of 0.837, compared to pNGAL (AUC 0.778) and serum creatinine (AUC 0.798) (Supplementary Figure S1).

Discussion

In this study of 110 patients with decompensated cirrhosis and AKI, we demonstrated that uNGAL could differentiate between ATN and other causes of AKI better than serum creatinine. uNGAL also showed moderate correlation with MELD score and CTP score on admission and was better than creatinine in predicting need for HD.

NGAL acts as a bacteriostatic agent in health, produced in many organs such as the kidneys, lungs, stomach, and colon.¹⁸ Circulating NGAL is filtered by the glomerulus and is reabsorbed in the proximal tubule and secreted into the thick ascending limb in low concentration.^19,20 After renal injury, the rise in uNGAL is due to both its increased expression in distal part of nephron as well as decreased reabsorption from proximal limb of loop of Henle and may rise as early as 2 h.^21,22

Although HRS is mostly considered as functional renal disease, histologic abnormalities such as tubular bile casts and tubulointerstitial injury have been described in small reports.^{23, 24, 25} Therefore, possible underlying tubular injury in HRS diagnosed by clinical criteria may explain raised uNGAL levels in our HRS patients. In addition, majority of our patients had ACLF, which is characterized by systemic inflammation. Since NGAL levels may also increase in inflammation, this may also be a factor contributing to its high level in the HRS subgroup. However, our study is not designed to evaluate this aspect, and further studies are required for the same. Concordant with prior studies, we showed uNGAL to be the highest in ATN followed by HRS and lowest in PRA.^{26, 27, 28, 29, 30} However, one patient in each HRS and ATN group had normal NGAL values despite AKI. We could not find any explanation for the same.

In a meta-analysis of five studies, the pooled sensitivity and specificity of uNGAL for the diagnosis of ATN were 0.86 and 0.82, respectively.³¹ We demonstrated a similar sensitivity, though the specificity was slightly lower in our study. Similar to a study by Belcher et al., we also demonstrated uNGAL to have higher diagnostic accuracy than serum creatinine in differentiating ATN and non-ATN AKI.²⁶ Early diagnosis of ATN and initiation of dialysis at a lower blood urea nitrogen level is associated with lower mortality and higher renal function recovery rates.^32,33 Since the initial management of AKI in cirrhosis involves plasma expansion for 48 h, earlier diagnosis of HRS may avoid unnecessary administration of fluids, which can precipitate noncardiogenic pulmonary edema.³⁴

uNGAL could predict the requirement of hemodialysis in our cohort of patients with greater accuracy than creatinine. The decision to hemodialysis was not based on NGAL values. Rather, the ability of NGAL to predict the need for hemodialysis was a finding of our study. We could not find any previous study that investigated this aspect. Therefore, patients with more severe renal injury can be identified early, and efforts can be intensified to salvage them. In future, studies are required that not only assess prospectively the ability of NGAL levels to predict hemodialysis but also assess the impact of such decisions on the mortality of patients with AKI.

Not many studies have looked simultaneously on urine and plasma NGAL values to differentiate between spectrum of AKI in cirrhosis. pNGAL levels are elevated in the presence of inflammation. The elevation is thought to be primarily due to release from neutrophils rather than renal tubules and has been shown to have a strong association with inflammatory cytokines. The weak correlation between pNGAL and creatinine observed in our study supports this hypothesis. Prior studies have shown pNGAL to be elevated in cirrhotics with AKI,³⁵ although they differed in its ability to differentiate between the types of renal dysfunction.^15,36 In our study, pNGAL could differentiate between the types of AKI and predicted the need for HD, although with lower accuracy than did uNGAL. This observation supports its association with systemic inflammation, which was present in majority of our patients.

In our cohort of patients, uNGAL was higher on admission in patients who died than in those who survived. It also correlated moderately with MELD score and CTP score. This is consistent with earlier observations that presence of AKI is associated with poorer outcomes in cirrhosis.^{30,31,35,37,38}

Our study is unique in few aspects. Most of the previous studies have evaluated NGAL in patients with decompensated cirrhosis with AKI. Our cohort consists predominantly of ACLF patients. AKI in ACLF differs from AD in that the underlying pathogenic mechanism is predominantly inflammation in the former.³⁹ Although NGAL has been found to predict need for HD in sepsis after cardiac surgery and coronary angiography in acute coronary syndrome, its ability to do so in cirrhosis has not been studied before.^{40, 41, 42} Our study should however must be interpreted in context of its limitations. The PRA group had lesser organ failure and was likely to have had lesser systemic inflammation than HRS and ATN groups. Lesser inflammation could have also contributed to the lower NGAL levels in that group. Inclusion of AD and ACLF made our cohort heterogeneous. Therefore, studies are needed that study uNGAL separately in these two populations. However, in our study ACLF patients were equally distributed in both HRS and ATN groups. Our cohort consisted of patients with established AKI, since our aim was to assess the utility of NGAL in discriminating between the various causes of the same. Future studies assessing NGAL levels at regular intervals in decompensated cirrhosis with normal creatinine may be useful in determining the ability of this biomarker in predicting the onset of AKI. Similarly NGAL at regular intervals in patients with AKI will also be helpful in determining the change in type of AKI during hospital course as AKI is a continuum. Some of our patients may have progressed from PRA to ATN or HRS to ATN, and this might have affected the conclusions regarding hemodialysis and death. However, we would like to point out that among the patients with PRA, none required hemodialysis. None of the 10 deaths in that group were related to renal failure. Of the patients with HRS, one developed shock and would have been excluded from definition of HRS. None received hepatotoxic drugs or diuretics or underwent repeat urine examination or NGAL estimation to look for microscopic hematuria or casts. Such tests would have posed financial constraints. So it is difficult to estimate how many with HRS subsequently fulfilled criteria for ATN. But since our secondary objective was to estimate diagnostic ability of NGAL to predict hemodialysis and mortality even if AKI subtype changed, it is unlikely to have affected the diagnostic ability of NGAL to do so.

In conclusion, uNGAL measures renal function better than serum creatinine and differentiates between structural and functional AKI with high accuracy. It is also a good prognostic indicator and correlates with conventional liver disease severity scores as well as predicts the need for hemodialysis in patients with decompensated cirrhosis. We suggest that uNGAL be incorporated into routine patient care in patients with cirrhosis.

Credit authorship contribution statement

Dr. Kshitiz Sharan—Manuscript preparation, data collection, and analysis.

Dr. Anand Sharma—conceptualization, Statistical analysis, and proof reading.

Dr. Satyavati Rana—conceptualization, data collection, and NGAL analysis.

Dr. Itish Patnaik—proof reading.

Dr. Rohit Gupta—proof reading and conceptualization.

Conflicts of interest

All authors have none to declare.

Funding information

None.

Appendix A. Supplementary data

The following are the Supplementary data to this article: